Most of the work is meant to happen without you sitting in front of the screen. Research Agent searches for strategy ideas, writes them as Java, and simulates them on market history, one strategy at a time. When a run finishes, the studio keeps the code, the profit, and the number of trades so you can compare versions.

The studio is the website after that search: the Strategy Editor, the Research Agent pages, Automations, and Support Desk. Fine-tuning a rule, testing it on another date range, and choosing where a live message should go are studio tasks. They are not done inside the Agent installer.

Live execution is also the website, not the Agent. The site watches closed bars and can send an event to a channel you configure. OpenTradeStudio never holds your funds and never sends an order to a broker. If an order is placed, it is placed by a script, a bot, or an AI agent that you run, under the terms of your broker.

Research Agent is a desktop program for Windows, macOS, and Linux. The installers are on the home page when a build has been published for that system. After install, the Agent opens its own page on that computer. On Overview you name the Agent and sign in with your OpenTradeStudio account. That sign-in registers the computer as your simulator.

The model stays on hardware you control. You can download an embedded model, which needs a graphics card with at least 8 GB of VRAM, or you can point the Agent at llama.cpp or Ollama already running on your network. Cloud AI is possible and is not the recommended setup: the Agent calls the model for hours, and about eight hours of those calls can spend a monthly token allowance.

The Agent works one strategy at a time. By default it spends about an hour on a strategy, then moves on, so the computer stays usable. That window is set in Configuration, from 15 minutes to 24 hours. When the window ends, a simulation that is still running is stopped.

Closing the browser tab does not stop the Agent. Use Quit in the tray. The Agent can update itself, or it can ask first. Enable and Disable for research live in the studio, on the Research Agents bar and in Configuration. Disable stops new research. A simulation you start yourself in the Strategy Editor can still run.

Strategy Editor

The buy and sell rules are Java on the left. The chart on the right draws the candles and marks buys and sells from a simulation. You can start from a template or open a strategy the Agent already wrote. Configure sets the market, the symbol, and the intervals. The assistant uses the same local model as the Agent: it explains the code and can edit it. A syntax check runs before a simulation. Your own strategies stay in My Strategies. Strategies the Agent creates stay in Agent Strategies.

Configuration

This page aims the Agent. Intervals start at 4 hours for a new agent, and at least one interval stays selected. Symbols are chosen by exchange. A document, a link, a YouTube video, or a note becomes source material for the next search. Limits set how many ideas may wait and how many minutes one strategy may take. The simulator fields on this page — date range, starting amount, and fees — apply to that agent’s research runs. They are separate from the simulator popup in the Strategy Editor. Enable starts research for that agent. Disable stops it.

Current activity

This is the strategy the selected Agent is working on now: the name, the improvement being written, and the profit and trade count so far. A clock shows how far the strategy is through its time window. If the computer is offline, the page keeps the last saved activity. One account can have several agents. The bar at the top of these pages chooses which agent you are looking at.

Research backlog

Ideas that have been collected and not run yet. They come from the web, from articles and YouTube, and from the documents and notes on Configuration. Each row is a written idea, oldest first. Star an idea to bring it forward, skip one to leave it out, or ask the Agent to analyse it next. A name box filters the list.

Simulated strategies and Top strategies

Simulated strategies lists what the Agent has already researched, newest first, with the best profit of each strategy. Open a run in the Strategy Editor to read the Java and test it on another range, or send it back to be analysed again. Top strategies ranks finished simulations that came out ahead. The profit on both pages is how the rules behaved on past candles. It is a way to compare versions. It is not a forecast.

Support Desk

Signed-in users open a ticket when something is wrong, when they have a question, or when they want a feature. Set a priority and a type, write what happened, and attach a file or paste a screenshot. Replies stay on the ticket.

Research can use crypto from Binance, US stocks and ETFs from Alpaca, and forex, metals, and oil from OANDA. Bars go down to one minute. Adding a symbol starts a history download. A simulation on that symbol waits until the history is ready. The website does not download the file itself; the connected Agent does.

The free Alpaca stock and ETF feed is delayed by about 15 minutes. A chart or a live signal on that feed can show a price that has already moved. Crypto, stock, and forex sessions are not the same. A feed can be incomplete or wrong, because it comes from another company. OpenTradeStudio does not correct a broker’s tape by rewriting it.

An automation is a strategy you chose, running on the website against live market data. Research Agent does not run it. When the rules say buy or sell, the site sends that event to the channels you turned on: Telegram, WhatsApp, Google Chat, Mail, Webhook, Discord, Slack, or Microsoft Teams. Send a test buy or a test sell before you start. The events list shows each signal and whether the alert arrived.

The message can be late, missed, duplicated, or wrong. Delivery depends on the network, on this website, and on the channel. The message is not an order. Your own executor decides whether to act, and the broker decides whether to accept the order.

While OpenTradeStudio is in Public Beta, Automations is free. There is no payment and no credit balance. An account can create up to five automations. A further create is refused. Research Agent and the studio stay free. When the beta ends, Automations may be charged at the price shown on the site before you confirm. That later price is presented on the home page as 49 US dollars a month.

A simulation describes the past. The Agent can try many ideas and keep the one with the best profit. That selection makes the number look stronger than a rule you had to choose in advance. A backtest can also assume a fill the live market will not give, and it may leave out the full cost of spreads, commissions, slippage, rejected orders, and gaps.

Use the studio to retest a candidate on another range and to compare profit and the number of trades. Do not treat a rank, a chart marker, or a story about another trader as a typical outcome or as advice. Trading stocks, ETFs, crypto, forex, metals, and oil can lose money. Where a broker allows leverage, a loss can be larger than the amount put in.